Personal agents are converging on the same few features
McBain (@heyneighbor) maps 47 personal agents (always-on AI assistants) as of November 5, 2026. Many now ship the same things: phone calls, checkout, chat apps and a one-tap yes.
This week · November 2–8, 2026
A weekly handful of the best things I read about AI.
McBain (@heyneighbor) maps 47 personal agents (always-on AI assistants) as of November 5, 2026. Many now ship the same things: phone calls, checkout, chat apps and a one-tap yes.
OpenUI Dashboards, from Thesys, lets an app's agent build dashboards inside it. After the model writes a short program from the app's own components and data tools, refreshes skip the model.
PageIndex, an open-source tool, builds a table of contents for each document instead of using vector search (matching text by closeness of meaning). A model reasons down it to the right pages and cites them.
Greg Isenberg of Late Checkout talks to agents more than to a 50-person team. The habits outlast his 30+ tools: grill each idea, gate every output and turn each mistake into a skill.
Slash, a business banking platform, says it passed $500m in yearly gross revenue with 88 people. Its CEO credits "SuperSMBs": tiny, fast-growing firms that buy on trust and want tools their agents can run.
A reducer is one function that turns a state and an event into the next state. dearlordylord writes a system's core as one, has the agent draw a diagram that replays real events, and asks why at each step.
Shopify's River agent works only in public Slack channels and declines direct messages. Staff learn by watching each other, so written fixes lifted its merge rate (code changes accepted) from 36% to 77% in two months.
AI-first small firms rebuild their go-to-market tools (the ones that find and win customers) with agents, says Anita Kirkovska, Vellum's head of growth. The CRM, the customer database, stays.
Ask an agent to make a page "better" and it adds a border. Impeccable, Paul Bakaus's open-source design skill, adds named moves, product context and a check for slop, the generic AI look.
McBain (@heyneighbor) maps 47 personal agents (always-on AI assistants) as of November 5, 2026. Many now ship the same things: phone calls, checkout, chat apps and a one-tap yes.
OpenUI Dashboards, from Thesys, lets an app's agent build dashboards inside it. After the model writes a short program from the app's own components and data tools, refreshes skip the model.
PageIndex, an open-source tool, builds a table of contents for each document instead of using vector search (matching text by closeness of meaning). A model reasons down it to the right pages and cites them.
Greg Isenberg of Late Checkout talks to agents more than to a 50-person team. The habits outlast his 30+ tools: grill each idea, gate every output and turn each mistake into a skill.
Slash, a business banking platform, says it passed $500m in yearly gross revenue with 88 people. Its CEO credits "SuperSMBs": tiny, fast-growing firms that buy on trust and want tools their agents can run.
A reducer is one function that turns a state and an event into the next state. dearlordylord writes a system's core as one, has the agent draw a diagram that replays real events, and asks why at each step.
Shopify's River agent works only in public Slack channels and declines direct messages. Staff learn by watching each other, so written fixes lifted its merge rate (code changes accepted) from 36% to 77% in two months.
AI-first small firms rebuild their go-to-market tools (the ones that find and win customers) with agents, says Anita Kirkovska, Vellum's head of growth. The CRM, the customer database, stays.
Ask an agent to make a page "better" and it adds a border. Impeccable, Paul Bakaus's open-source design skill, adds named moves, product context and a check for slop, the generic AI look.
DeepSeek's Harness team designed one always-on agent that runs many tasks at once. A router talks to people while isolated workers do the tasks.
The stack-diagram skill by @caru_ini turns one sentence about a tech stack into an architecture diagram with official logos. The agent renders each diagram and looks at it before handing it over.
Agents can build faux UI (mock product screens for a marketing page), but it often looks generated. Designer Benji Taylor's Stageshow skill has the agent copy a polished example, then film its own work and fix it.
Talking beats typing for speed and for context. Anomaly makes the OpenCode agent and ships Hex, a free dictation app that runs on your own computer.
A text box wired to a coding agent turns each request into a shipped change. Stripe CEO Patrick Collison puts one in every personal app and calls it "liquid software."
Cloudflare, once a content delivery network, now covers compute, databases, file storage and background jobs. Coralie of Railblocks calls it an easy place for AI agents to build and ship apps.
Coding agents had repos, tests and tickets to work in. Lorenzo Castro says operations agents need an operating system that gives each job its own context, tools, state and access.
Every company rents the same frontier models, so the lasting edge is a record it owns: its data, definitions, permissions and graded results. Each correction fed back in teaches the next model what the last one learned.
Ben Horowitz and Databricks CEO Ali Ghodsi say agents fail on missing context, not intelligence. With one record of facts, meanings, rights and results, people and agents need fewer managers.
Jev, a model from TypeSafe AI, answers set questions with a typed result: a label, a score or a yes/no probability, plus a confidence. Paired with Firecrawl's company records, it sorted 146 startups by who buys from them in 2.2 seconds.
Ibo (@ibothinks) swaps Next.js, Vercel and Supabase for Cloudflare's own parts, on one bill from $5 a month. It fits a new small app, but the database and the bill need care.
Firecrawl's free Government & Legal Index searches 30 million US government pages, from federal statutes to city permits. Each result quotes its source and the laws it cites, so a person can check the answer.
Marcel Kargul built a polished CRM (customer-tracking tool) screen in two days with Claude and two design skills. It has no database, sign-in or API behind it.
Vercel automates dull, repetitive work by first shadowing the person who does it best. Its lead agent researches sales leads and drafts replies, and a person approves every send in Slack.
Last year's model handles most back-office work, says Varick's Daniel Kornum. Agent misses come mostly from missing context and bad data, not the model.
An agent that repeats a mistake is missing a written rule. Hamza Khalid wrote one per failure of a personal agent, starting with a margin quoted from eight-day-old data.
Agent-team failures often come from handoffs, not the model. Nizam's design concept OrchestrateIQ checks each handoff like a contract and tests fixes on past runs.
TasteSkill v2 by Leon Lin gives a coding agent an art direction, real photos and a browser check of its own page. Its blind test found the check removed flaws but didn't raise scores.
Mike Taylor, co-author of Prompt Engineering for Generative AI, treats an eval (a repeatable test of AI work) as saved work replayed on each new model or skill. Grading is 90% of the job.
Wispr Flow, a dictation app valued at $2 billion in August 2026, gets 500 to 700 new ads a month from about 50 paid creators. Each creator covers a customer gap no other ad reaches.
Anatomy is an open-source skill that turns a one-line idea into an interactive 3D-style drawing of a machine. A cache key that skips repeat work becomes a lock that turns only when every pin matches.
Cycle.js creator André Staltz says three moves cover most UI design. Claude Design (Anthropic's design tool) drafts, an agent critiques the draft against real app screens from Mobbin, and a person with taste cuts.
A short prompt leaves every design choice in an AI motion video to chance. As of November 2026, p4n's Prompt Motion gallery shows over 230 Opus 5.5 videos beside the prompts that made them.
Thariq of Anthropic's Claude Code team built html-plan, a plugin that writes a plan as one interactive page. Each claim shows one proof, such as a mockup or call stack, and each decision sits on the claim it changes with Claude's pick checked.
A non-engineer can now get a custom part made: measure, have a model write CAD (computer-aided design) code, print a plastic test, then order from an online factory. Greg Isenberg's 13-step playbook answers Nikita Bier, who ordered a wifi-locked dog door with…
Andrej Karpathy expects people to spend more time understanding model output. He ranks formats by ease: plainer English, diagrams, web pages, then custom explainer videos.
Claude Code's advisor lets the main model ask a stronger model for review at key moments. @mirku21 pairs it with cheaper Sonnet subagents (helper agents) for routine work.
Vox (@Voxyz_ai) makes motion videos in Claude Code with four roles: breakdown, storyboard, build and review. HeyGen's open-source HyperFrames turns HTML scenes into video.
Claude Code's playground plugin builds a one-file web page with controls, a live preview and a prompt to paste back. Thariq at Anthropic uses it where text falls short.
Claude Code can build an eval (test cases and a grader), then hillclimb: change one thing at a time and keep what scores better. Gains count only on cases Claude never studied.
Clay's AI writing policy, now for all 550 staff, lets AI help draft. The author must stand behind every line, because readers' time counts more than the writer's.
In September 2026, investor Sam Lessin showed one Claude Code engine behind every door, from email to about 140 scheduled jobs. All share one trust check, memory and model router.
Effort sets how hard Claude works on a task. Thariq of Anthropic's Claude Code team found it mostly buys testing.
A new model pays off only after the old prompts are cleaned up. Lance Martin of Anthropic paired the Opus 5.5 launch with a command for each step.
Thariq used Claude Code workflows (scripts that run many subagents) to have Opus 5.5 redesign a personal website many times and critique each round. A cheaper top model made it affordable.
Tamara Tran's plugin compacts a Claude Code session (frees room in its context) by deleting old tool calls and results instead of summarizing. TypeSafe's Jev model picks what goes.
Claude Code Projects turns one chat into a coordinator that runs parallel cloud sessions, called threads. Every thread reads the shared goal, so setup time belongs there.
Browser Use's open-source Jev Ultrafast turns each web page into a numbered list of controls. TypeSafe's Jev model picks one option per step instead of writing text, so a flight search took 7 seconds.
Polar is an AI browser whose agents work for hours on any site as the logged-in user. Kevin Jiang (@jiang_kev) announced its $5.7M seed round in July 2026.
Most PDFs already carry their text, so OCR (reading a page as an image) is often wasted work. Firecrawl's open-source pdf-inspector checks a PDF in about 20 ms, extracts the text as Markdown locally and sends only text-free pages to OCR.
Lewis Carhart of compliance startup Comp AI built a CRM (customer tracking tool) in half a day instead of paying HubSpot $80,000 a year. The open-source version centers on an agent that records only what it can prove.
JJ Englert (@JJEnglert) found Claude Opus 5 chatty, overconfident and tiring at its July 2026 launch. A short reply style section in the global CLAUDE.md (the file Claude Code reads in every project) tamed it.
Anthropic cut over 80% of Claude Code's system prompt for its Claude 5 models and lost nothing on its coding tests. Newer models need judgment and short files that point to detail.
Models overuse "corrective juxtaposition": "the answer isn't X, it's Y." Vox (@Voxyz_ai) adds six CLAUDE.md rules that cap it and other sentence-shape tics.
The best model changes weekly, so Ramp routes each request to the cheapest model that clears a quality bar. Ramp's in-house router opened to US users as Router in August 2026.
Banning words like "delve" one by one keeps the same AI voice. Vox (@Voxyz_ai) pastes Orwell's six 1946 writing rules into a global CLAUDE.md instead.
Skills, plugins and always-loaded instructions pile up and cost context (the model's working memory) in every session. Claude Code's checkup finds that waste and fixes it once you say yes.
Peter Steinberger's codex-first skill saves metered Claude tokens. Claude keeps design, judgment and final review, and Codex workers do the hands-on coding.
A loop is an agent repeating a cycle of work until a stop condition is met. Anthropic's Claude Code team sorts loops by what starts and stops them.
Hundreds of hours of trial and error, handed over for free. Even if you rewrite half of it, a working reference library beats a blank page every time. Go take what's useful.
A cost pattern worth stealing. Let the expensive frontier model do the planning and hand the grunt work to cheaper ones. Obvious in hindsight, and it quietly changes how you budget a project.
The clearest writing yet on why the model stopped being the bottleneck. Thariq's frame: the gap between what you asked for and what the work actually needs is where quality leaks out, and the real skill is surfacing those gaps before, during, and after you build. Worth reading twice.
shadcn shipping composable components for chat UIs, starting with the streaming conversation layer. Dry on its own, but it matters: as every product bolts on a chat surface, having clean, ownable primitives beats reinventing streaming and message state each time. One for the builders.
A heavy dictation user, top 0.1%, ditched his paid voice tool for an open-source one that runs locally on his Mac with no API key and handles slang better. The broader signal worth watching: local, private, free tools are quietly getting good enough to cancel the subscription. Short and practical.
Anthropic's playbook for working alongside agents as the tools ship into Slack. The interesting shift is treating an agent less like software you run and more like a teammate you delegate to, which changes how you hand off and review work. Read it before your org figures this out the hard way.
A tight framing of the shift everyone's circling: you stop prompting the agent and start building the system that prompts it. Schedule, discover, build, verify, repeat. Worth it as the one-line mental model before you read anything longer on loops.
The whole thing is one habit: every day, Claude builds and prioritizes a to-do list against your actual goals, then calendar-blocks it. Deliberately unimpressive, which is the point. The skills that stick are the boring ones you run every morning, not the clever ones you run once.
The announcement: Claude now joins Slack as a taggable team member for Enterprise and Team plans. Worth noting because Slack, not a terminal, is where most teams already coordinate, and putting the agent there changes who ends up using it. The reference point for the human-agent-teams conversation that followed.
Thariq's field notes on Claude-in-Slack, a genuinely new form factor that nobody has best practices for yet. Early and speculative, but that's the value: you're reading someone figuring out the patterns in real time rather than after they've calcified. Good companion to the launch post.
A pointer to Addy Osmani's walkthrough of how the software lifecycle is changing, phase by phase, from autocomplete to autonomous agents. Osmani is a reliable, non-hypey guide to this stuff, so it's a solid framework read if you want the shape of the shift rather than a hot take.
Boris Cherny on living inside Artifacts: visual explanations of tricky code, system diagrams, dashboards he shares with the team, all built from the session at a private link. The signal here is a power user telling you which everyday habit actually changed how he works. Small feature, big workflow shift.
Andreessen's long, unusually good essay reframing SpaceX as the one company assembling the full stack for a post-scarcity future: cheap launch, orbital compute, lunar industry, Mars. The engineering-culture sections carry it, especially Musk's five-step Algorithm and the idiot index. Not AI-specific, but the orbital-data-center thesis ties directly to where compute and energy are heading. A real sit-down read.
Block's numbers on their internal agent system are the ones to quote in an argument. Engineers tag it in Slack, it researches, plans, and ships: 200,000 operations a day, 1,500 PRs merged a week, 15% of all production code. This is what agents coordinating across a real codebase looks like at scale, not in a demo.
A launch pitch dressed as a provocation: AI is making marketers lazy, so make the website do the work instead. Mostly a product announcement with a $27M seed behind it, but the underlying bet, that the site itself becomes an active agent rather than a brochure, is where a lot of marketing tooling is heading.
A neat idea made real: a feedback loop that catches visual slop and design-system drift while the agent builds, instead of after. It turns design consistency from a thing you ask for into a hook that runs automatically. If your agents produce frontends that look slightly off every time, this is the missing guardrail.
A shared full guide to shipping your first agent, with the pitch that it collapses a two-week slog into a day. Standard on-ramp material, but a decent starting point if you're going zero to one and want a single path instead of ten browser tabs.
Karpathy's line, put to work: stop being the bottleneck, put in few tokens, have a huge amount happen on your behalf. The framing is that loop engineering is the mechanism that actually delivers that. A quick nudge toward the mindset shift, not a how-to.
The best long read on why a top-tier model isn't the point. The system around it is. Loops, memory, verifier sub-agents, and state files are what make each run leave the next one smarter, and the piece lays out the whole stack with cost-routing advice for when to reach for the expensive model versus a cheap one. Long, but it's the map most people are missing.
Steinberger's minimal recipe: an orchestrator that wakes every five minutes and directs work to threads, combined with triage, auto-review, and computer-use skills so work lands on its own. The takeaway is how little scaffolding a useful loop actually needs once you have sharp skills to call. Concrete and copyable.
The demo that makes the capability click: on a customer call, Claude transcribed in the background and built the features the customer was wishing for in real time, ready by the end of the call. One anecdote, but it reframes what fast means. This is the moment a lot of people realized the ground had moved.
A non-engineer spent a billion tokens running Claude's top model on real work across writing, strategy, security, and design. Refreshing because it's a user's honest read on where the model actually delivers, not a benchmark chart. Useful counterweight to the launch-day hype.
An Anthropic engineer on two things the top model changes: self-correction loops and memory. The sharpest takeaway is that the model shouldn't grade its own work. An independent verifier explores harder and recovers from dead ends where self-critique stalls at good enough. Short, concrete, and from someone who actually ran the experiments.
The best explainer on the phrase everyone was repeating without defining. A loop is cron plus a decision-maker in the body: the model, not a hardcoded script, picks the next move each tick. The punchlines land, that the loop, not the model, is now the expensive part, and that the real asset is the skills a loop calls, not the loop itself. If you read one thing on loops, this.
A plain-English tour of the Claude features hiding in plain sight: Projects, memory, extended thinking, scheduled tasks, prompt caching, custom roles. Nothing exotic, but the value is in the framing. Each one takes minutes to set up and pays off daily. Good to forward to anyone still treating Claude as a fancier search box.
Just a pointer to a GitHub repo of agent skills. Low signal on its own, but these skill collections are where the real how-to lives. Clone it, read how the skills are structured, keep what fits.
Anthropic's own writeup on building agents that do data analysis. The useful parts are the unglamorous ones: skills, clean data foundations, and evaluations. If you're wiring an agent into real business numbers, this is the part everyone skips and then regrets.
The clearest map of where interfaces are heading: agents drawing the UI in real time instead of describing it. Three patterns, controlled, declarative, open-ended, each with a different failure mode at scale, and most teams pick one by accident. If you build anything agent-facing, this is the decision tree to read before you're locked in.
A clean way to think about agent design: the model is deliberately thin, and intelligence gets pushed outward into memory, skills, and protocols that the harness composes at runtime. The useful question it hands you is where any new capability should live. Good conceptual scaffolding if agents still feel like a bag of tricks.
An honest field report: dynamic workflows felt useless until they were pointed at adversarial review. The move is splitting a review into thin, mean, single-focus lanes, correctness, duplication, safety, each agent required to bring a way to verify its own finding. It beats self-review because the model likes its own work too much. The step-by-step is copy-and-adapt ready.
The line most people skim: every agent in a workflow can run a different model. This is the proof. A 33-agent audit, cheap readers plus one smart synthesizer, ran in under five minutes and found a real buried error. Put the expensive model where the thinking happens and run the rest on the cheap one. That's the economics of parallel agents in one example.
The canonical piece on dynamic workflows, from the Anthropic engineer who built them. Claude writes its own custom harness on the fly to beat the failure modes of one long context window: laziness, self-preference, goal drift. The example prompts alone are worth the read, and it names the reusable patterns (fan-out, adversarial verify, tournament) you'll see everywhere else. Start here.
A short thread on a real problem: when agents do the work, how do you actually understand what got done? The value is in seeing how people inside Anthropic keep oversight without babysitting. Worth a scroll if your agents are starting to outrun your attention.
A design studio's field report on a real shift: the deliverable is no longer a PDF brand guide but a folder of structured files an agent can build from. The value moves upstream to the thinking, and the sharp bit is what they call magic_trick.md, the one human, left-of-center idea the system can't generate on its own. Best essay here on where human creativity stays scarce.
A sharp look at how youth sports became a $40 billion industry pricing out ordinary families. Not an AI piece, but a clear read on how private capital reshapes a market that used to be a childhood pastime. Worth it if you care about where consolidation goes when it meets kids and weekends.
A pointed demo of where the moat isn't: someone cloned the core of two heavily funded legal-AI products in two weeks and open-sourced it. The argument, that firms should own their application layer instead of renting it forever, is the uncomfortable question every vertical SaaS company is now facing. Provocative, and hard to dismiss.
Karpathy's follow-up carries the sharper point: in the agent era you stop shipping code and start shipping the idea, kept deliberately vague, and the other person's agent builds it for their needs. It's a small reframe with big implications for how software gets distributed. The knowledge-base spec is the worked example.
Karpathy on a workflow more people should steal: point an LLM at a pile of raw sources and let it compile and maintain a markdown wiki you rarely touch by hand. Once it's big enough, you query it like a research assistant, no fancy RAG required. The best part is that your own explorations file back in, so the knowledge base compounds. One of the most-shared AI ideas of the year for a reason.
A genuinely good step-by-step on running Claude Code 24/7 on an always-on Mac Mini so you can text it tasks from anywhere. The best detail is the one nobody mentions: there's no message queue, so a sleeping laptop drops everything, which is the whole case for dedicated hardware. Practical if you want a personal agent that never goes dark.
A crisp reminder of what belongs in your project config. Most hand-written setup was stuff the agent can discover by reading the code. The new approach keeps only what Claude would get wrong without it: security rules, mandatory workflows, non-obvious gotchas. Delete the rest. 458 lines to 68.
The reference for anyone configuring Claude Code seriously. It walks the whole control center: CLAUDE.md, path-scoped rules, and hooks, with the key nuance that instructions are suggestions but hooks are deterministic. Keep CLAUDE.md under 200 lines or adherence drops. This is the one to bookmark.
No reads match. Try another topic or search.